针对光场深度估计过程中数据量大、边缘处深度估计结果不准确问题, 利用压缩感知原理重建光场, 提出一种新的多信息融合的光场图像深度估计算法。利用压缩感知重建算法重建5×5视角光场数据, 获取光场数据后首先移动子孔径实现重聚焦, 然后利用角度像素块散焦线索和匹配线索计算出场景初始深度和置信度。计算图像边缘信息, 通过融合初始深度、置信度、边缘信息获取最终深度。实现压缩光场仿真重建, 并对仿真光场数据和公开光场数据进行深度估计, 实验结果表明: 可以仿真重建出5×5视角光场数据, 且仿真重建的光场可用于深度估计。该深度估计算法在场景边缘处的深度估计结果边界清晰, 层次分明, 验证了重建光场深度估计的可行性与准确性。

Aiming at the problem of large amount of the light field data and inaccurate depth estimation at the edge, the light field was reconstructed by compressive sensing principle and a depth estimation method based on multiple information fusion was proposed . Firstly, a 5×5 view light-field was reconstructed according to the compressed sensing reconstruction algorithm, and the refocusing was realized by moving the sub-aperture after obtaining the field data. Then, the initial depth and confidence of the scene were calculated by using the defocused response cues and correspondence response of angular patch. Finally, the image edge information was calculated, and the final depth was obtained by fusing the initial depth, confidence and edge information. The compressed light field simulation and physical reconstruction were realized, and the depth estimation of the simulated reconstruction light field and the open optical field data were also achieved. The experimental results show that the method proposed can simulate and reconstruct the 5×5 view light field data, which can be used for depth estimation. The depth estimation results of the proposed method at the edge of the scene is clear and hierarchical, which verifies the feasibility and accuracy of the depth estimation for the reconstructed light field based on compressed sensing theory.